The antecedents of work performance have long been under scrutiny. Transformational leadership (TFL) positively predicts employees' work performances (Ceri-Booms et al., 2017; DeGroot et al., 2000). The team/organization level can not be neglected when analyzing the relationship between TFL and performance (Braun et al., 2013). Recent research about attachment theory in the workplace demonstrated the importance of attachment in affecting job performance (Richards & Schat, 2011). Leaders perceived as attachment figures contribute to organizational outcomes (Molero et al., 2019). In their systematic review, Yip and colleagues (2017) identified the need to go beyond the "typical" adult attachment styles applied to organizational research. They recommend revisiting the basics of attachment theory and focusing more on attachment figures and their functions that create the preconditions for attachment relationships. The current study aims to analyze the functions of a leader as an attachment figure in predicting job performance via TFL. For the first time, we would like to verify the attachment relationships at the team level through multi-level structural equation modeling (multi-level mediation analysis).

Transformational leader and attachment figure

Leadership behavior is typically clustered into two main orientations, task- and person-focused behavior (Fleishman et al., 1991). TFL is thought to fall into the second one (Burke et al., 2006). Transformational leaders influence their followers via charisma, inspirational motivation, intellectual stimulation, and individual approach (Bass & Riggio, 2006). Transformational leaders are focused on individuals, and they are attentive to the needs of their followers. Such leaders can easily be viewed as leaders with attachment figure characteristics.

Attachment-based relationships are typically described using leader characteristics such as supportive, available, responsive, sensitive, or referring to the quality of leader–follower relationships (Fein et al., 2020). Attachment figures have five basic functions. (Simpson et al., 2020). They create emotional ties with the attached individual. Attachment figures provide a secure base (by being supportive, encouraging, responsive, and available) and a safe haven (by being protective, reassuring, and soothing). They respond to proximity-seeking from the attached individual. Separation from the attachment figure can elicit separation distress in the attached person.

The central tenet underpinning attachment theory is the secure base concept of the attachment figure (Waters & Waters, 2006). Attachment figures provide a secure base by being supportive, encouraging, responsive, and available. Further, attachment figures provide a safe haven to their followers by being protective, reassuring, and soothing. Mayseless and Popper (2019) propose a care and competence model for leaders to enable them to fulfill their followers' needs. The model explicitly highlights leaders' secure base and safe haven functions of attachment figures. Transformational leaders encourage followers to perceive them as attachment figures. Attachment security is formed through the image of ourselves and of others. Others are expected to be available, responsive, and supportive (Bowlby, 1973). This is enhanced by repeated positive interpersonal interactions with attachment or significant others (Mikulincer & Shaver, 2016). Mikulincer and Shaver (2020) described the process as a broaden-and-build cycle of attachment security. Differences in attachment dynamics are attributed to the quality of these interactions in times of need (Mikulincer & Shaver, 2016). A security-enhancing leader is defined as a sensitive and responsive leader who can support and empower followers and create a broaden-and-build cycle of attachment security. Secure-base support encapsulates availability, encouragement, and non-interference in the followers' work (Feeney & Thrush, 2010). The leaders' secure-base support is beneficial particularly for followers with insecure attachment (Wu & Parker, 2017).

Another function of the attachment figure refers to separation distress. Followers who suffer from separation distress want to keep physically close to the leader and not let them go. A non-security-providing leader could be described as insensitive to followers' needs, unavailable, or inconsistent in supporting followers, and therefore elicits the activation of attachment strategies. Consequently, followers become too focused on their attachment needs and respond with hyperactivating or deactivating attachment strategies (Yip et al., 2017). Hyperactivating attachment strategies entail overdependence on others and having a low view of oneself. Deactivating strategies entail independence from others and a high view of oneself. Separation distress can be an ongoing stressor even when the actual stressor has receded. Using a hyperactivating strategy in attachment results in a vicious self-amplifying cycle of distress, in which negative emotions escalate through ineffective behavior (Mikulincer & Shaver, 2016).

The recent literature defines transformational leaders as attachment figures in theoretical terms (Mayseless & Popper, 2019). Several studies empirically examined a follower's attachment to a leader (Fein et al., 2020; Molero et al., 2013, 2019). Molero and colleagues (2019) developed the Leader as a Security Provider Scale to cover attachment figure functions in leaders. Their scale investigates a specific-relationship attachment model to a specific leader (Mikulincer & Shaver, 2016; Sibley & Overall, 2010). Although the scale is believed to tap all functions of the attachment figure in leaders, it is unidimensional. However, later studies have shown two dimensions of the scale, "secure figure" and "separation distress" (Lisá et al., 2021; Mrázková & Lisá, 2022).

Based on those mentioned above, we can assume that TFL affects the perception of leaders as attachment figures.

Transformational leaders and performance

The relationship between transformational leadership and performance is well documented (DeGroot et al., 2000). Work performance is usually described as both an individual's direct and indirect contributions to an organization's goals (Borman & Motowidlo, 1997). The scientific community has a broad consensus that individual work performance consists of two interplaying components. They are task (in-role) and contextual (extra-role) performance. The first refers to formal job performance; the second is organizational citizenship behavior. In their three-cluster model of citizenship performance, Coleman and Borman (2000) differ interpersonal citizenship performance (cooperative and facilitative efforts), organizational citizenship performance (commitment to the organization), and task conscientiousness (extra efforts to maximize job performance). Apart from the obvious importance of task/in-role performance, organizational citizenship behavior is also highly welcome by employers (Lisá et al., 2019).

TFL encourages emotional attachment to the organization (Thompson et al., 2021), organizational commitment (Rodríguez-Fernández et al., 2021), and extra-role behaviors (Pradhan & Pradhan, 2016). DeGroot and colleagues (2000) reported a modest effect size of TFL on followers' commitment. Braun et al. (2013) demonstrated that TFL also enhances performance at the team level. Meta-analyses show that TFL significantly relates to different levels of performance. It relates more to organizational than individual-level performance (DeGroot et al., 2000). Empirical knowledge shows small to moderate relationships between TFL and team performance (Ceri-Booms et al., 2017; DeGroot et al., 2000).

These results led us to build the current study model as a two-level model. It seems appropriate to explore TFL using multi-level analysis (Bureau et al., 2021; DeGroot et al., 2000; Gagné et al., 2020). We can assume that TFL predicts performance at the team level.

Attachment figure and performance

Many studies have established the importance of attachment at work (Grady et al., 2019; Harms, 2011; Mikulincer & Shaver, 2016; Yip et al., 2017). For example, the global or domain attachment related to leader-member exchange (Fein et al., 2020), performance (Davidovitz et al., 2007), social exchanges (Richards & Hackett, 2012), organizational citizenship behavior (Reizer et al., 2021; Richards & Schat, 2011), job satisfaction, and burnout (Kafetsios et al., 2014). These studies examined leaders' and/or followers' dispositional attachment orientation. They applied instruments primarily designed for romantic relationships or relationships with significant others.

Less is known about the leader as a complex attachment figure and their role in predicting work performance. Various partial findings regarding attachment characteristics of leaders are reported. Leaders' support is a crucial job resource for organizational outcomes (House, 1971). Perception of leaders as supportive one related to followers' job satisfaction, job commitment, burnout, and individual performance assessed by achieving set goals (Lavy, 2014), proactive behavior (Wu & Parker, 2017), organizational citizenship performance (Lisá et al., 2021; Molero et al., 2013), extra and in-role performance (Shanock & Eisenberger, 2006).

Attachment figures possess complex characteristics, and only a few studies acknowledged their multiple functions. Molero and colleagues (2019) found that leaders as complex attachment figures enhance employee work outcomes such as engagement, job satisfaction, and organizational identification. Moreover, they found that TFL leadership enhances the psychological safety climate and prevents employee job burnout. Lisá and colleagues (2021) found that secure-figure-providing leaders enhance loyalty and adherence to organizations' rules in engaged followers. On the other hand, the perception of separation distress invokes ignoring rules and neglecting organizational goals.

We can conclude that the leader's functions as an attachment figure can affect the performance of followers.

Research goal

The current study will contribute to attachment theory in the workplace and expand the literature on TFL. Leaders as attachment figures may produce different performance outcomes predicted by TFL. We will include various types of performance in the analysis (task/citizenship). Therefore, we will differentiate the mediation effect of perceiving a leader as an attachment figure (security provider) on predicting various types of performance by TFL.

Second, we incorporate both leader and follower perspectives. Leaders evaluated their leadership qualities and their team's performance. Followers assessed their general/extra-role performance and their leaders as attachment figures. We will contribute to the leader-oriented leadership literature by considering the role of followers, who constitute a crucial group in leadership models (Chen et al., 2018). This multi-source approach helps reduce common method bias (DeGroot et al., 2000).

Third, by using the specific research tool (Molero et al., 2019) designed for the analysis of perceiving a leader as an attachment figure, we reflect on the need to examine relationship-specific attachment. Relationship-specific attachment is better suited to research in work and organizational settings than the general attachment model. Leaders in a relation-specific attachment may not be not a "significant other" or "romantic partner" as is usual in the general attachment models. We also consider whether relation-specific measures should be used (Fraley et al., 2011) and to what extent they can help in explaining and predicting organizational outcomes. The results of the specific-relationship model will extend the literature on attachment work models (Mikulincer & Shaver, 2016).

We will conduct multi-level analysis because leaders display different relationships towards individual followers and groups (DeGroot et al., 2000). We, therefore, examine data at individual and team levels and thus provide complex solutions for understanding work performance through attachment figures. The current study analyzes two mediating variables yet not considered (Secure figure and Separation distress as characteristics of attachment figure in leader), including two levels of analysis (individual and team level) and data from leaders' and employees' perspectives. The current study is the first one that links various measures of predicted performance (in- and extra-role performance, team performance) with dynamics of perceiving a leader as a security provider. The multi-level approach is typical for analyzing TFL, but not the attachment. We will contribute to the attachment theory in the workplace from a multi-level perspective.

We assume that the perception of leaders as attachment figures (reported by followers) would mediate the relationship between TFL and performance (reported by the leader or follower). Both leaders and followers are causal agents influencing performance (Chen et al., 2018). Therefore, we included their views in predicting performance and used multi-level structural equation modeling.

We examine attachment figures functions and use Molero and colleagues' (2019) scale of perceiving the leader as a security provider. Although the scale is constructed to tap all functions of the attachment figure in leaders, it is unidimensional in the original study. However, later studies have shown two dimensions of the scale, "Secure figure" and "Separation distress" (Lisá et al., 2021; Mrázková & Lisá, 2022). Therefore, we acknowledge this two-factorial solution and transfer it into the following hypothesis. Figure 1 reflects the common complex model for hypotheses H1-H5.

  • H1: Dimensions of perceiving a leader as a security provider (secure figure and separation distress assessed by followers) will mediate the relationship between transformational leadership (assessed by leader/followers) and team performance (assessed by leaders).

  • H2: Dimensions of perceiving a leader as a security provider (secure figure and separation distress assessed by followers) will mediate the relationship between transformational leadership (assessed by leader/followers) and general performance (assessed by followers).

  • H3: Dimensions of perceiving a leader as a security provider (secure figure and separation distress assessed by followers) will mediate the relationship between transformational leadership (assessed by leader/followers) and interpersonal citizenship performance (assessed by followers).

  • H4: Dimensions of perceiving a leader as a security provider (secure figure and separation distress assessed by followers) will mediate the relationship between transformational leadership (assessed by leader/followers) and organizational citizenship performance (assessed by followers).

  • H5: Dimensions of perceiving a leader as a security provider (secure figure and separation distress assessed by followers) will mediate the relationship between transformational leadership (assessed by leader/followers) and task conscientiousness (assessed by followers).

Fig. 1
figure 1

Complex mediation model with two parallel mediators. X = TFL_L, M1 = SF, M2 = SD. Y = WPT (for model A), Y = GP (for model B), Y = CP1 (for model C), Y = CP2 (for model D), Y = CP3 (for model E)

Methods

Participants

Participation in the research was anonymous and voluntary, based on convenience sampling. To contact the leaders, we use the snowball method. We asked the leaders to collect data from their teams for the individual feedback report. The participants submitted their completed questionnaires directly to the researchers, so the leader did not see their responses. The participants completed the questionnaires using pen-paper self-administration. Since the data collection was anonymous, we do not know the percentage of team members participating in the research.

The research sample consisted of 294 participants, of whom 239 were employees, and 55 were their leaders. Five teams had two leaders. We added to each participant both leaders with different team numbers. After excluding cases with missing values, teams with less than three observations, and outliers, the research sample had 204 employees nested in 38 teams. The 204 employees (44% men, 49% women, and 7% item non-response) were aged from 19 to 63 years (AM = 38.7 years; SD = 9.6). They had worked with their current leader from 0 to 35 years (M = 5.6, SD = 5.5). Thirty-eight leaders of the work teams (55% men, 37% women, and 8% item non-non-response) were aged 19 to 62 years (AM = 43 years; SD = 10) and led their teams from 1 to 30 years. Leaders managed teams of three to twelve, with five members on average. The leaders and employees worked in various market sectors.

The leaders and employees completed separate questionnaire batteries. The questionnaires completed by leaders were considered level 2 and by followers level 1 measure. The work teams were given identification numbers. The team members fulfilled questions about perceptions of their leader as a security provider, transformational leadership, and general and citizenship performance. The leaders answered questions about their transformational leadership style and work team performance.

Power analysis

There are some recommendations about the sample size in multi-level structural equation modeling (MSEM) in the literature. A sample of N = 200 observations provides a reasonable estimate for CFI and TLI when the number of observed variables is less than 30 (Shi et al., 2019). For MSEM, forty clusters are a minimal requirement for large path detection at the team level (Meuleman & Billiet, 2009). In contrast, a sample with fewer than 20 clusters increases estimates bias. In organizational MSEM research, we can find studies with a comparable number of clusters, like 34 (Chen et al., 2017) or 39 (Braun et al., 2013), but with a bigger number of participants (360–423).

The power analysis for MSEM showed a power of 0.85 for two-level analysis with three arms, 38 clusters with five members, and 0.17 ICC on average (Zhang and Yuan, 2018). To our best knowledge, there is no commonly available tool for power analysis for multi-level mediation analysis. But we computed it for mediation analysis (Schoemann et al., 2017). Power analysis for mediation with two parallel mediators uses a Monte Carlo simulation. It tested the indirect effect with a 90% bootstrapped confidence interval (Schoemann et al., 2017). It only showed enough power (0.92, 0.87) for the ab path in models with TLF_L as an exogenous variable and CP2 and CP3 outcomes.

We used different sources for the estimation of power analyses. We showed that the current sample with 204 participants and 38 clusters is on the threshold of needed statistical power. There is a risk that the results will not reveal the medium or small estimates, and we will not have enough statistical power to deny the null hypothesis.

Study instruments

The questionnaires were translated into Slovak by two independent translators. The researchers obtained a final version via consensus. The aim was to reflect the intended meaning (Greškovičová, 2020).

The Leader as Security Provider Scale measures perceptions of the leader as an attachment figure (Molero et al., 2019). The original scale consisted of 15 items rated on a five-point Likert scale. These ranged from 0 (strongly disagree) to four (strongly agree). We applied the Slovak translation (Lisá et al., 2021) and the two-factorial scale model. The "Secure Figure" factor (SF) included seven items (α = 0.895; ω = 0.90). An example item is "I can count on my leader to support me when I propose new ideas or procedures." SF expresses the idea that an attachment figure provides a secure base and safe haven. Employees tend to perceive their attachment figure or leader as supporting and encouraging their pursuit of non-attachment goals. They tend to perceive their leader as a source of protection, comfort, calmness, and reassurance when needed (Molero et al., 2019). The "Separation Distress" (SD) factor included three items (α = 0.801; ω = 0.810), for example, "If my leader moved to another organization or another position in this organization, I would try to go with him/her." SD measures the intensity of how employees build emotional ties with their leaders and feel distressed when separated from their leaders and try to stay in their leaders' proximity. The confirmatory factor analysis (CFA) showed acceptable data fit for the two-factor model [(X2 = 68.881(34), p < 0.001; TLI = 0.995, CFI = 0.996, RMSEA = 0.065, and SRMR = 0.053)].

Czech Leadership Questionnaire measures TFL through 16 items (Procházka et al., 2016). The leaders assessed their TL (TLF_L) on a seven-point Likert scale from 1 (never) to 7 (always), for example "I keep my promises" (α = 0.939; ω = 0.940) and so did the followers (TFL_F; α = 0.946; ω = 0.947). The CFA for TLF_L [X2 = 131.889(104), p = 0.034; TLI = 0.995, CFI = 0.996, RMSEA = 0.073, and SRMR = 0.081] and for TLF_F [X2 = 241.061(df = 104), p ˂ 0.001; TLI = 0.995, CFI = 0.995, RMSEA = 0.078, and SRMR = 0.052] showed acceptable data fit. The TFL of the questionnaire consists of the following dimensions: idealized influence, inspirational motivation, intellectual stimulation, and individual consideration. We used the recommended one-dimensional second-order factor model (Procházka, 2020).

The General Work Performance Questionnaire (Motowidlo & Van Scotter, 1994) measures perceived general work performance. It has three items on a seven-point Likert scale. Seven indicates a high performance and one a low performance. Questions: "My team exceeds/ does not meet standards for job performance," "My team performs poorly/high compared to other teams in the same area," "My team compared to other teams in the company, it contributes poorly/more to the effectiveness of the organization." The leaders assessed the performance of their team (WPT). The mean of the three items formed the overall performance score. The internal consistency for leader-assessed team performance was α = 0.905; ω = 0.906. The followers assessed their general performance (GP). The internal consistency for followers perceived performance was α = 0.801; ω = 0.802.

Citizenship Performance (Coleman & Borman, 2000) measures citizenship organizational behavior. The scale consists of 23 items scored on a 5-point Likert scale, from 1 (never) to 5 (always). It includes three subscales: CP1—interpersonal citizenship performance as assisting, supporting, and developing co-workers through cooperation beyond expectations (8 items, α = 0.852; ω = 0.853); CP2—organizational citizenship performance as demonstrating commitment to the organization through loyal complying with rules, policies, and organizational objectives (9 items, α = 0.843; ω = 0.856); CP3—task conscientiousness as extra efforts that go beyond role requirements, and dedication to the top achievements (6 items, α = 0.783; ω = 0.787). Fit indices of the scale: X2 = 444.853(206), p = ˂ 0.001; TLI = 0.985, CFI = 0.986, RMSEA = 0.075, and SRMR = 0.072.

Ethical considerations

All participants in this study provided informed consent in writing. All the current study procedures conformed to the ethical standards of the 1964 Helsinki declaration and its later amendments and Internal Institutional Regulation Nb. 23/2016. FSEV UK ethics committee declared, under Statement No. FSEV823-2/2022SD that no formal ethics approval was required in this particular case.

Data analysis

We processed the data using R 4.2.1 and JASP 0.16.3.0. We performed a descriptive analysis, reliability analysis, Pearson correlation analysis, confirmatory factor analysis (with 5000 bootstrap replications), and multi-level mediation analysis. The model fit indices for CFA were: the comparative fit index (CFI), Tucker-Lewis index (TLI), standardized root mean of residuals (SRMR), and root mean squared error of approximation (RMSEA) (Byrne, 2009; Kline, 2005). Dataset, syntax, and Appendix are available at Figshare with DOI number https://doi.org/10.6084/m9.figshare.19747435.

We computed a multi-level analysis in R 4.2.1 lavaan (Rosseel, 2022), using estimator MLR, which is maximum likelihood estimation with robust (Huber-White) standard errors and a scaled test statistic. In path analysis, we simultaneously fit individual and team models. ICCs of all individual-level variables crossed the 0.05 cutoff (Table 2), which indicates that the multi-level techniques for regression models may be appropriate and profitable (Preacher et al., 2010).

Firstly, we analyzed 20 single models (with one mediator) through multi-level mediation in 2–1-2, 2–1-1, 1–1-2, and 1–1-1 designs. In the 2–1-2 design, the model included the exogenous variable TFL_L, mediator SF/SD at level 1, and outcome WPT at level 2. In the 2–1-1 design, the model included the exogenous variable TFL_L, mediators SF/SD at level 1, and outcomes GP, CP1, CP2, and CP3 at level 1. In the 1–1-2 design, the model included the exogenous variable TFL_F, mediators SF/SD at level 1, and outcome WPT at level 2. In the 1–1-1 design, the model included the exogenous variable TFL_F at level 1, mediator SF/SD at level 1, and outcomes GP, CP1, CP2, and CP3 at level 1. We tested the fit of 20 single mediation models (see Appendix, table A1). The models with TLF_L, an exogenous variable, showed zero in chi-quadrat and degrees of freedom; we did not interpret the fit indices. The models with TLF_F, an exogenous variable, did not fit well with the data (see Appendix, table A1 at https://doi.org/10.6084/m9.figshare.19747435). In a complex mediation model with both mediators, we estimated only models with TFL_L as an exogenous variable. The complex models with two mediators had the 2–1-2 and 2–1-1 designs (Fig. 1).

The multiple data sources' condition helped reduce a potential common method bias (Podsakoff et al., 2012; Rodríguez-Ardura & Meseguer-Artola, 2020). Although in both designs, there were two variables from the same level. In the 2–1-2 model, leaders assessed the exogenous variable and the outcome (level 2). In the 2–1-1 model, followers assessed the mediators and the outcome (level 1). Altogether we tested five complex models, A to E.

Results

No variables showed significant gender differences except SD and CP3, but with a small effect size. The correlations at the individual level among age, time spent with the leader, and the measured variables (Table 1) were low. We concluded that demographic variables did not affect the rest of the measured variables. SF (individual level) correlated the most with TLF_F (team level) (r = 0.609***) and TFL_L correlated the most with team performance (r = 0.534***, both at the team level). The factors of perceiving a leader as a security provider and the factors of citizenship performance correlated moderately to large, which confirms the closeness of the constructs' dimensions. The individual level variables included in mediation analysis were normally distributed; therefore, we did not mean centered them.

Table 1 Descriptive Statistics and Correlations for Study Variables

Models A to E had zero for chi-square and degrees of freedom, so we did not interpret their fit indices. Table 2 shows the unstandardized coefficients. The perception of a leader as a security provider did not affect how the leader assessed the team's performance (Model A). The secure figure predicted GP at level 1 (Model B). Separation distress predicted CP1 positively at level 1 and negatively at level 2 (Model C). The Secure figure predicted CP2 positively at level 2 (Model C). The Secure figure predicted CP2 positively at levels 1 and 2. Separation distress predicted CP2 positively at level 1 and negatively at level 2 (Model D). Secure figure and separation distress predicted CP3 only at level 1 (Model E). None of the indirect effects were significant.

Table 2 Unstandardized coefficients of the multi-level models

Discussion

The study aimed to analyze the mediation effect of perceiving a leader as a security provider on performance predicted by TFL. First, the simultaneous models with perceived TFL as an exogenous variable did not fit the data well. Because of that, we only computed models with TFL reported by the leader (exogenous variable). Second, the TFL of leaders predicted dimensions of secure leaders significantly and positively at the team level in all tested models. Third, we did not support any mediation hypotheses at the team level (H1-H5). However, we found some remarkable partial results that we would like to discuss further.

Team performance and general performance

The dimensions of perceiving a leader as a security provider (secure figure and separation distress) did not contribute to the team performance prediction. How the leaders assessed their teams did not relate to how followers perceived them as security providers. However, at the team level, TFL positively predicted followers' perception of leaders as security providers (Model A). Model B with general performance as outcome showed similar results. Perceiving a specific leader as a security provider in followers did not contribute to their perceived task performance (at the team level). Similarly, previous research denies the relationship between the constructs (Lisá et al., 2021). The possible interpretation has to do with the nature of the task performance. Task performance focuses more on the task than on the relationships. However, attachment is defined explicitly in terms of relationships to an object. Harms (2011) also adds that citizenship behavior refers to discretionary helping behaviors that are likely to happen when a relationship is formed. We need to note that our results contradict some previous studies (Lavy, 2014; Shanock & Eisenberger, 2006). Based on stated contradictions, we assume task performance might be differently approached and operationalized. Therefore the relationship between attachment and performance results are divergent.

Citizenship performance

Separation distress predicted interpersonal citizenship positively at the individual level (Model C). But the prediction turned out to be negative at the team level. On the other hand, the secure figure did not predict interpersonal citizenship performance at the individual but did at the team level. When followers perceive a leader as a secure base and safe haven, they tend to cooperate with other team members. A previous study also corroborated this result (Lisá et al., 2021).

The secure figure positively predicted organizational citizenship performance at the individual and team levels (Model D). When followers perceive leaders as secure figures, they tend to comply with rules, policies, and organizational objectives. This was also supported by previous research (Lisá et al., 2021). This finding is corroborated by other research results from specific workplace samples. For example, trust in leaders fostered better safety compliance in healthcare (Enwereuzor et al., 2020). Support from leaders improved safety compliance in the food service industry (Guchait et al., 2016). Conversely, a lack of support weakened safety compliance in the navy (Kjellevold Olsen et al., 2021). The knowledge that the perception of a leader as a secure figure can contribute to loyal compliance with rules is important. In the ongoing health crisis, there is a greater emphasis on safety measures and employee compliance (Hu et al., 2021).

The separation distress positively predicted organizational citizenship performance at the individual and negatively at the team level (Model D). The direction of its prediction went opposite to the TFL direction. Separation distress at the individual level contributed to organizational citizenship performance, but on the team level, it goes against it. It means that when followers are distressed and experience hyperactivating strategies, they seek proximity with the leader. Followers are thus demotivated to demonstrate commitment to the organization. For them, a leader represents the whole organization. We can assume that the leader means the entire organization to followers with high separation distress towards their leader. When followers fail to maintain close contact with the leader, they lose connection to the organization.

The secure figure and the separation distress positively predicted task conscientiousness at the individual level but not at the team level (Model E). Task conscientiousness relates to extra effort in achieving working goals. The leader, as an attachment figure, has the potential to motivate achieving the operational objectives at the individual level of employees' perception.

Individual vs. team level of analysis

Dimensions of a leader perceived as a security provider predicted the perceived performance of followers more often at the individual than at the team level. It seems that perceiving leaders as attachment figure differs between individual and team levels. The relationship dynamic with a leader as an attachment figure appears to differ from that relationship at the team level. The relationship between transformational leadership and performance via the leader as an attachment figure is more valid at the individual level. When followers' unique perception of an attachment figure aggregates, the individualized perceptions are lost, and the effects may be insignificant. People usually do not share their experience of an attachment figure in groups. The experience of an attachment figure is not cumulative but individual. The individual follower-centric approach of the leader may be essential in a one-to-one relationship. The leader's relevance as an attachment figure may attenuate at the group level.

We can explain it with the separation distress example. Separation distress, positively predicted by TFL, predicts citizenship performance positively at the individual but negatively at the team level. The separation distress dimension in the current study is typical of the desire to be in a leader's proximity. Hyperactivated separation distress dimension means that people want to be in touch with their leader, despite the quality of an emotional relationship and the situation. They do not want to be without a leader. They are willing to quit the job when the leader leaves. The proximity is easier to maintain at the individual level of the relationship. At the team level, maintaining proximity to the leader is much more challenging. If a follower lacks proximity to the leader, the hyperactivation attachment system can become active (Yip et al., 2017). The employee is under stress that relates to the worse perceived citizenship performance. There could be exceptions. The work position of personal assistants could be one where the perceived separation distress would not disrupt the citizenship performance.

On the other side, perceiving a leader as a secure figure predicted interpersonal citizenship behavior only at the team, not at the individual level. Similarly, several studies show individual and team-level differences in transformational leadership. In their meta-analysis, DeGroot and colleagues (2000) reported similar results when TFL affected followers less at the individual than group level. They concluded that transformational leaders should exhibit similar and consistent behavior towards each group member. When the behavior differs towards the group members and follows the principles of TFL (individual approach), the effect drops at the individual level. The distinct behavior, tailored, individually focused, and malleable, decreases the effect. There are significant between-group differences because the followers perceive their leaders individually (Yammarino et al., 1993). Another multi-level analysis (Bureau et al., 2021) showed that the relationship between transformational leadership and incivility differs at individual and group levels. The constructs were evaluated differently regarding the individual/group level. Finally, in their meta-analysis, Gagné et al. (2020) also specified the difference between the perception of controlled motivation at the individual and group levels. Collective perceptions of transformational leadership related to increased collective autonomous and controlled motivation. Individually perceived controlled motivation related to increased individual perceptions of transactional leadership.

In the current study, there are two distinct dimensions of perceiving the leader as an attachment figure: secure figure and separation distress. These dimensions encompass the different functions of an attachment figure. Therefore, they differ in their consequences for work settings. Moreover, they relate to contrasting aspects of an attachment figure regarding what is required and desirable in leaders (Mayseless & Popper, 2019). Furthermore, the functions of the attachment figure can be expressed differently at the individual and team levels. It could be also helpful for leaders in the role of attachment figures to distinguish between individual and team level communication with the individuals/teams.

The limitation and future research implications

The limitation of the study lies in the small, borderline number of participants and clusters included in the MSEM. The possible explanation of why the indirect effects at the team level were insignificant may lie in the small statistical power of the research sample (Schoemann et al., 2017). At least the mediation models with interpersonal and organizational citizenship behavior deserve to be examined on a larger sample because of all significant paths (ab) in the mediation models. Although we used the multi-source of data, which is considered the prevention against common method bias, the same data source in the case of followers (models 2–1-1) or leaders (model 2–1-2) could potentially cause a common method bias. To be sure, the next research could apply a collection of the predictors and outcomes at different points in time, in different locations, or using different media (Rodríguez-Ardura & Meseguer-Artola, 2020).

The current study showed more significant performance predictions at the individual than at the team level. It seems appropriate to look for other team variables that affect the perceiving a leader as a security provider at the team level. We hypothesize that some unknown team variables affect (maybe moderate) the relationship between secure leaders and performance outcomes.

In the current study, leaders assessed their TFL. They presented their view of themselves and how they meet the criteria of transformational leadership. They generally reported on how they act, not relate to individuals, but how they behave in leading their team. The followers assessed how they perceived the leader subjectively as an attachment figure. A leader as an attachment figure is a very subjective individual construct. It is a result of a specific-relationship model which depicts a specific relationship. This view is highly individualized, and each team member perceives the leader-attachment figure differently. Some followers perceive leaders as sensitive, supportive, and available, some as insensitive to employees' needs, unavailable, or inconsistent. This individualized view also relates to follower attachment, as found in previous research (Davidovitz et al., 2007; Wöhrle et al., 2015). We demonstrated that transformational leadership positively predicts dimensions of perceiving a leader as a security provider. In future research, it would be valuable to explore the options for influencing the perception of a leader as an attachment figure through leaders' TFL and by leaders themselves.

We did not explore the interaction between the global or domain attachment styles of followers and their perception of the leader, which could provide more complex insights into the processes between leaders and their followers (Keller, 2003). Another direction for future research could be to look at attachment dynamics in various types of the work team, e.g., collocated, virtual, or hybrid teams. The use of such teams has become more prevalent since the onset of the COVID pandemic. For instance, by looking more closely at team activities and the leader as a secure figure, we may also be able to better explain cognitive or emotional overload in avoidant individuals (Mikulincer & Shaver, 2016). Another research path would be to explore within-fluctuations of attachment representations (Girme et al., 2018), following day-to-day manifestations of the attachment figure and perceptions of it in the work setting. Moreover, because anxiety declines and avoidance increases in older adults more than in younger adults (Chopik et al., 2013), it would also be interesting to examine age and its role in specific-relationship models.

We applied convenient sampling, so the results are not representative. A homogenous sample (for example sample of personal assistants) or a sample from one organization might produce different results. Followers may hold biases about their leaders and may overestimate or underestimate them depending on their attachment orientation (Mikulincer & Shaver, 2016). We applied a cross-sectional approach that disabled us from interpreting the causal effects. A longitudinal study could show how attachment figure perception in employees causally relates to transformational leadership and its outcomes over an extended period (Mullen & Kelloway, 2009). Another limitation concerns the number of team members. We could ensure that every team member participated, so we have limited perceptions of team members.

We could not analyze the mediation models with TFL perceived in followers because of the bad fit indices of single mediation models. The question is if the model itself has a meaning or if this is a dead end.

Practical implications

Feeney and Trush (2010) propose to use attachment towards the specific objects to predict attitudes and behaviors towards these objects. This would mean that we can predict attitudes and behaviors towards the leaders through attachment to them. We broaden their view finding evidence that attachment to leader widens its range of prediction beyond the attitudes and behaviors towards leader.

Repeated positive experiences between leaders and followers promote attachment security. They attenuate stress levels, improve adaptive coping, sustain resilience, and lead to positive mental health outcomes (Feeney & Fitzgerald, 2022; Simpson et al., 2020). In the current study, separation distress negatively affected performance at the team level. Followers who suffer from separation distress want to keep physically close to the leaders and not let them go. Therefore, followers need some awareness of an attachment figure to understand what leader support is and what it relates to. Followers would benefit from identifying the symptoms of separation distress so they can mitigate it. Separation distress can relate to feelings of anxiety. Elevated employee anxiety reduces organizational compliance (Rahmani et al., 2021). Situations of perceived danger, such as taking on a new responsibility, may prove risky for employees with high perceived separation distress. Followers' perception of threats in the workplace may be a fascinating area to address concerning the broaden-and-build cycle of attachment security (Mikulincer & Shaver, 2020).

The leader as an attachment figure does not operate in the same way as a parental figure. The two dimensions of secure figure and separation distress are aligned with four criteria of attachment figures in adult relationships (Simpson et al., 2020): secure base, safe haven, proximity seeking, and separation distress. Secure figure encompasses secure base and safe haven functions that seem salient in work and organizational settings, as indicated initially by the care and competence model of Mayseless and Popper (2019). On the other hand, the separation distress dimension encompasses proximity seeking and separation distress functions. Separation distress goes against interpersonal and organizational citizenship at the team level. Therefore, a strong desire for physical proximity between leaders and their followers that goes beyond the professional relationship does not bring added value to companies. Close emotional relationships as function attachment figures may even hinder the leader and follower's work performance (Lanaj & Jennings, 2020).

The current study's findings may help leaders understand how employee perceptions of them can affect followers' performance. Being a transformational leader evoking security or separation distress helps to modify leadership behavior with better outcomes. Security-enhancing (Fein et al., 2020) and transformational leadership (Khan et al., 2020) training help increase the perceived citizenship performance at the team level. Leaders should enhance followers' security to improve work performances in crises and ordinary situations (Grady et al., 2019).

The contribution of the study

Applying attachment theory to work and organizational settings raises exciting questions regarding using global or specific attachment models (Mikulincer & Shaver, 2016; Sibley & Overall, 2010). The results show that the specific leader–follower relationship model is worth further investigation using a specially tailored tool, like the Leader as a security provider scale (Molero et al., 2019).

The construct of the leader as an attachment figure is well described in theoretical works (Mayseless & Popper, 2019; Mikulincer & Shaver, 2020). The empirical construct for measuring it was developed recently (Molero et al., 2019). This is the first time perceiving a leader as an attachment figure is analyzed on the team level. The results show that the team affects the perceived leader as a security provider. This knowledge moves the subjectively-based construct of attachment to the new team-level context. Although attachment can be in the direct connection to the performance at the individual level, in the context of the team, this connection can be modified.

We applied the two-sourced data, which helped us to reduce common-method bias. Moreover, the simultaneous analysis is sensible for building the model in complexity, including all theoretically motivated random effects (Preacher et al., 2010).

Conclusion

The current study aimed to analyze the relationship between transformational leadership and performance with a leader as an attachment figure as the mediator. We confirmed a two-dimensional solution for the leader as an attachment figure (secure figure and separation distress). The perception of a leader as an attachment figure did not significantly mediate the prediction of followers' work performance by leaders' transformational leadership. Perceiving the leader as a secure figure positively predicted the interpersonal and organizational citizenship performance at the team level. Separation distress negatively predicted interpersonal and organizational citizenship performance at the team level but positively at the individual level. Perceiving a leader as an attachment figure was a more often significant predictor of performance at the individual than team level. The perception of a leader as an attachment figure did not significantly predict task performance at the team level. At the team level, transformational leadership positively predicted the perception of the leader as an attachment figure. The findings suggest that leaders contribute to their followers' performance as attachment figures. As seen through the lens of transformational leadership, the best scenario for team performance is a secure-figure leader (a secure base and safe haven) and followers with low separation distress.